I read with interest the article entitled "Effectiveness and Cost-effectiveness of Albendazole in improving nutritional status of pre-school children in Urban slums"(1). I would like to make the following observations on this article.

This article claims to be single blinded, placebo controlled trial with child as the unit of randomization. However, the randomization performed is questionable. The target of the study was to include 500 children in each of the two arms. However, out of a total of 1061 eligible children, they enrolled 610 subjects in the Albendazole group and the remaining 451 children in the placebo group. This is mainly because the authors relied on the serial numbers allocated by the Anganwadi workers for randomization. Those with odd or non-zero endings numbers were assigned to placebo group and rest to the intervention group. As we understand, there are certain randomization procedures (e.g., block randomization or simple randomization procedures) which necessarily deal with atleast the first unit to be randomly allocated to either of the group. This is not true in the present study in view of anticipated presence of systematic bias present in the listing of the children. Hence, I would consider that this study has been carried out by purposive sampling method instead of claimed random sampling method. If investigators could have used even a simple randomization procedure, they would have got almost an equal number of children in each arm of the study resulting in a high power of the study. Further, in the absence of proper randomization, it is very difficult to maintain the comparability of the groups in relation to even known variables, which is evident from Table I. For example, history of worms is more in placebo group (4.9%) in comparison to Albendazole group (2.6%). Similarly, measles vaccination cover-age at enrollment was also found more in placebo group (89.6%) in comparison to Albendazole group (84.8%). In absence of proper randomization any clinical trial looses its merit as well as utility.

This paper, further, claims that Albendazole reduces the risk of stunting. However, Table I, clearly reveals that average height gain of placebo group is slightly more (mean = 10.3 cm; SD = 51 cm) in comparison to Albendazole group (Mean = 9.94 cm; SD = 4.9 cm). Further, if we look at the weight gain and height gain in both the groups, they are statistically comparable. However, if we look at the figures given in Table II, there is a decrease in proportion of underweight children as well as wasted children and an increase in proportion of stunted children in both the groups at the end of the study. Use of an inappropriate method of comparison of results (e.g., Chi-square test) may have resulted into significant results. Here, it would have been more appropriate if McNemars’ Test was used for comparison of pre and post trial status in terms of proportion of under-weight, stunted and wasted children. Since, these results are on the same group of children presented, results in terms of per cent difference may give us erroneous results. For example, per cent difference shows decrease in proportion of underweight/wasted children but an increase in proportion of stunted children. Hence, I feel that these results should not have been included in Table II. Proportion of stunted children in both the groups clearly reveals that there is an increase of stunted children even in Albendazole group at the end of the study which is contrary to the results in Table I showing comparable height and weight gains in both the groups. It is very difficult to agree with the observations made by the authors under the conclusion in absence of appropriate analysis. One may be interested to see propor-tion of the children who became underweight, stunted and wasted during two years of study period alongwith those who have overcome these problems during same period. Further, if we look at the results given under Table III (A ) as well as III (B ), Relative Risk is such that there have been similar changes in both the groups at two years. Also both the groups had comparable incidence density per hundred child years for various infections (Table IV ). Thus, these results also do not indirectly support the conclusion.

It has been concluded that Albendazole proved effective in reducing the risk of stunting and incremental cost effectiveness has been calculated. However, an important aspect of analysis was overlooked by the investigators. There was no significant differ-ence in the height gain in the two groups. Infact, the height gain in the placebo group was marginally higher (10.35 cm vs 9.94 cm). The study had enough sample size to detect a mean difference of height gain of 1 cm with 90% power or 0.85 cm with 80% power in the two groups at 5% significance level. What then is the logic for suggesting Albendazole as a cost effective intervention for dramatically reducing the risk of stunting when it is obvious that the height of the population will remain unchanged. It may have been desirable to depict cost effectiveness of Albendazole for improving the height of population by 1 cm or more if it had been statistically significant.

Umesh Kapil,
Additional Professor,
Human Nutrition Unit,
All India Institute of Medical Sciences,
New Delhi 110 029,
India.

1. Awasthi S, Pande V, Fletcher RH. Effectiveness and cost-effectiveness of albendazole in improving nutritional status of pre-school children in urban slums. Indian Pediatr 2000; 37: 19-29.

I would like to thank Dr. Kapil for raising these issues as it has given me another apportunity to clarify certain points as well as correct an error in the published article.

Let me explain the method of randomization. The list of eligible children was prepared by the project workers by abstracting data from the register maintained by the anganwadi worker. The sequence order of the names was not changed. This list was brought to the office and the children were given a serial number by project staff different from those who went to the field. The staff in the first two anganwadi centers allocated zero ending numbers as well as other odd ending numbers, that is 1,3,5,7,9, to the albendazole group entirely by mistake. When this was discovered, the project consultants were immediately informed. They advised that this procedure be continued to avoid introduction of systematic bias. It would very minimally decrease the power of the study without influencing the effect, the alpha level for a two tailed p value of <0.05 which was considered statistically significant. Hence the albendazole group ended up having 60% and placebo group 40% of children. The simplest but valid randomization procedure was adopted and it is incorrect to call this purposive sampling. The typographical error highlighted below may have led Dr. Kapil to undermine the randomization process.

In the main paper, a typographical error was overlooked during the proof reading on page 20 column 2, line 10 should read "Those with even and non-zero ending numbers ..... rest to the intervention group". The correction has been underlined.*

Randomization is a technique for con-trolling for confounders but it is by no means foolproof. Many randomized trials have not been able to control for all the confounders, some of which are as basic as sex. This is common knowledge. As required, we have reported those variables where randomization could not correct imbalances, like proportion with history of worm infestation and measles immunization. This adds to the credibility of the results and the trial certainly does not loose its merit as a result of this.

The height gain in Table I in the alben-dazole and placebo groups is not statistically significant. Once appropriate testing has been done what is the value of commenting on average height gain? This can mislead a reader. It is strange that in one place a comment is made on the inappropriateness of Chi-square statistical test while in the same letter "average" height gain is being given importance.

It is well known that the proportion of stunting in low income countries like ours increases with increasing age in the first five years. Our data is consistent with this. We have used Chi-square test for comparison of cate-gorical variables in our study. The test is perfectly appropriate. As the cell size was <1000, Fisher’s exact was used. McNemar test is basically used in calculation of Chi-square test in matched paired studies with 2 or more comparison groups; no matching of subjects was done in the two arms of the current trial.

Stunting is not synonymous with cessation of linear growth. Therefore children can gain suboptimal weight or height and be categorized as underweight or stunted, respectively. If the deficit is greater in height gain than in weight gain, the proportion stunted would exceed the proportion underweight. Let me also point out that Tables IIIA and B are for comparing the effect of deworming on the cognitive functions, using R-PDG assessment, and not for assessing impact on nutritional status. Table IV shows incidence density of infections. Lack of demonstrable difference in infection episodes does not invalidate the results on stunting.

Mortality and poor cognitive outcomes have been related to stunting in children. Stunting is a measure which compares incident height to reference height after controlling for age and sex. If a higher proportion of the population is not stunted in childhood there may be a reduction in mortality rates and improved cognitive functioning in later life. Are we interested in our children just being taller or having better chances of survival into produc-tive adulthood? Hence, the cost effectiveness for the prevention of stunting has been calculated.